FitLocally by GPUwerk

Qwen · 27.3B

Can you run Qwen3.8-27B locally?

Qwen3.8-27B is a dense model with 27.3B parameters. At 4-bit or better it runs on 74 of the 120 hardware configurations listed here; the smallest is the RX 7900 XT (20 GB). On a 128 GB DGX Spark it runs at Q4_K_M, at roughly 11 tokens per second (reported).

Data as of 2 November 2026

Fastest machines for Qwen3.8-27B

Every machine that runs it, fastest first. Filter by type; your machine is marked. Compare machines running it side by side

  1. 1RTX PRO 6000 BlackwellGraphics card · 96 GB~55 tok/s ModelFit est.Q8_0 · 29 of 95 GB
  2. 22x RTX 5090Graphics card · 64 GB~54 tok/s ModelFit est.Q8_0 · 29 of 62 GB
  3. 32x RTX PRO 6000 BlackwellGraphics card · 192 GB~54 tok/s ModelFit est.Q8_0 · 29 of 190 GB
  4. 4RTX 4090Graphics card · 24 GB~51 tok/s ModelFit est.Q4_K_M · 17.9 of 23 GB
  5. 5RTX 3090 TiGraphics card · 24 GB~48 tok/s ModelFit est.Q4_K_M · 17.9 of 23 GB
  6. 6RTX 3090Graphics card · 24 GB~45 tok/s ModelFit est.Q4_K_M · 17.9 of 23 GB
  7. 7RX 7900 XTXGraphics card · 24 GB~44 tok/s ModelFit est.Q4_K_M · 17.9 of 23 GB
  8. 8RX 7900 XTGraphics card · 20 GB~37 tok/s ModelFit est.Q4_K_M · 17.9 of 19.0 GB
  9. 9RTX 5090Graphics card · 32 GB~36 tok/s our est.Q8_0 · 29 of 31 GB
  10. 10M5 UltraMac · 96–512 GB · 5 variants~34 tok/s ModelFit est.Q8_0 · 29 GB
  11. 112x RTX 4090Graphics card · 48 GB~30 tok/s ModelFit est.Q8_0 · 29 of 46 GB
  12. 122x RTX 3090 TiGraphics card · 48 GB~29 tok/s ModelFit est.Q8_0 · 29 of 46 GB
  13. 132x RTX 3090Graphics card · 48 GB~27 tok/s ModelFit est.Q8_0 · 29 of 46 GB
  14. 142x RX 7900 XTXGraphics card · 48 GB~27 tok/s ModelFit est.Q8_0 · 29 of 46 GB
  15. 15M5 Max · 36 GB · 32cMac · 36 GB~23 tok/s ModelFit est.Q4_K_M · 17.9 of 24 GB
  16. 16M3 UltraMac · 96–512 GB · 5 variants~22 tok/s ModelFit est.Q8_0 · 29 GB
  17. 17M2 UltraMac · 64–192 GB · 7 variants~21 tok/s ModelFit est.Q8_0 · 29 GB
  18. 18M4 Max · 36 GB · 32cMac · 36 GB~19 tok/s ModelFit est.Q4_K_M · 17.9 of 24 GB
  19. 19M1 UltraMac · 64–128 GB · 4 variants~19 tok/s ModelFit est.Q8_0 · 29 GB
  20. 20M2 MaxMac · 32 GB · 2 variants~19 tok/s ModelFit est.Q4_K_M · 17.9 GB
  21. 21M5 MaxMac · 48–128 GB · 3 variants~18 tok/s ModelFit est.Q8_0 · 29 GB
  22. 22M1 MaxMac · 32–64 GB · 4 variants~17 tok/s ModelFit est.Q4_K_M · 17.9 GB
  23. 23M5 ProMac · 48–64 GB · 4 variants~17 tok/s ModelFit est.Q4_K_M · 17.9 GB
  24. 24M4 MaxMac · 48–128 GB · 3 variants~15 tok/s ModelFit est.Q8_0 · 29 GB
  25. 25M3 MaxMac · 36–96 GB · 2 variants~15 tok/s ModelFit est.Q4_K_M · 17.9 GB
  26. 26M4 ProMac · 48–64 GB · 4 variants~14 tok/s ModelFit est.Q4_K_M · 17.9 GB
  27. 27Ryzen AI Max+ 395 · 128 GBAI mini PC · 128 GB~12 tok/s ModelFit est.Q4_K_M · 17.9 of 96 GB
  28. 28M3 MaxMac · 48–128 GB · 3 variants~12 tok/s ModelFit est.Q8_0 · 29 GB
  29. 29DGX Spark (GB10)AI mini PC · 128 GB~11 tok/s reportedQ4_K_M · 17.9 of 122 GB
  30. 30M2 MaxMac · 64–96 GB · 3 variants~11 tok/s ModelFit est.Q8_0 · 29 GB
  31. 31M2 Pro · 32 GBMac · 32 GB~10 tok/s ModelFit est.Q4_K_M · 17.9 of 21 GB
  32. 32M6 · 32 GB · 12cMac · 32 GB~9.8 tok/s slow ModelFit est.Q4_K_M · 17.9 of 21 GB
  33. 33M1 Pro · 32 GBMac · 32 GB~9.2 tok/s slow ModelFit est.Q4_K_M · 17.9 of 21 GB
  34. 34Ryzen AI Max+ 395 · 64 GBAI mini PC · 64 GB~8.5 tok/s slow our est.Q4_K_M · 17.9 of 48 GB
  35. 35M5Mac · 32 GB · 2 variants~8.3 tok/s slow ModelFit est.Q4_K_M · 17.9 GB
  36. 36M3 Pro · 36 GBMac · 36 GB~8.1 tok/s slow ModelFit est.Q4_K_M · 17.9 of 24 GB
  37. 37M4 · 32 GBMac · 32 GB~6.1 tok/s slow ModelFit est.Q4_K_M · 17.9 of 21 GB
Too big for 46 machines

M5 Pro · 24 GB · 20c, M5 Pro · 24 GB · 16c, M4 Pro · 24 GB · 20c, M4 Pro · 24 GB · 16c, M6 · 24 GB · 12c, M6 · 24 GB · 10c, M5 · 24 GB · 10c, M5 · 24 GB · 8c, M4 · 24 GB, M3 · 24 GB, M2 · 24 GB, RTX 5080, RTX 5070 Ti, RTX 5060 Ti, RTX 4080 SUPER, RTX 4080, RTX 4070 Ti SUPER, RTX 4060 Ti, Tesla T4, Tesla P100, RX 9070 XT, RX 9060 XT, M3 Pro · 18 GB, RTX 5070, RTX 4070 Ti, RTX 4070 SUPER, RTX 4070, RTX 3060, M2 Pro · 16 GB, M1 Pro · 16 GB, M6 · 16 GB · 12c, M6 · 16 GB · 10c, M5 · 16 GB · 10c, M5 · 16 GB · 8c, M4 · 16 GB, M3 · 16 GB, M2 · 16 GB, M1 · 16 GB, RTX 5060, RTX 4060, RTX 3070, RTX 3060 Ti, RTX 2070, M3 · 8 GB, M2 · 8 GB, M1 · 8 GB. Qwen3.8-27B needs about 17.9 GB at its smallest build.

Speeds are single-user decode estimates, not measurements. "ModelFit est." comes from ModelFit's model fitted on published runs (CC BY 4.0); "our est." is memory bandwidth divided by bytes read per token, times 0.55. Memory is what the best build needs, against what the machine can give a model. Method

DGX Spark, reported

11.4 to 24.7 tok/s

range across quantizations and engines. Source and setup in the GPUwerk guide

About the model

Parameters
27.3B
Active per token
27.3B
Type
Dense
Good for
Strong general dense model
Model card
Hugging Face
Hugging Face count
27.78B (agrees)

Builds and how to run them

QuantizationLoadMin. RAMCommand
Q4_K_M16.5 GB24 GBollama run qwen3.8:27b
Q8_027 GB48 GBollama run qwen3.8:27b-q8_0

Load, minimum RAM and Ollama commands: ModelFit, CC BY 4.0. Load is memory with the model loaded, before long contexts.

Questions

Can I run Qwen3.8-27B on a Mac?

Yes on a Mac with enough memory: the smallest Apple configuration listed here that runs it at 4-bit or better is the M2 Max · 32 GB · 38c.

How much memory does Qwen3.8-27B need?

About 17.9 GB at the smallest build in our data, including a margin for the runtime and an 8K-token context. Longer contexts need more.

Does Qwen3.8-27B run on a DGX Spark?

Yes, at Q4_K_M. The Spark has 128 GB of unified memory, about 121.6 GB usable.